Hi Ingo, The specific case of a single-element ndarray which has a badvalue returning true in a Boolean context is clearly a bug, and will be fixed (https://github.com/PDLPorters/pdl/issues/388). But unfortunately reverting the whole change would now (half a year after) involve breaking other code which has been adjusted accordingly (some of which I now maintain), so that would not now make things better, including that the old behaviour is genuinely surprising. Sorry.
Best regards, Ed From: Ingo Schmid<mailto:ingo...@gmx.at> Sent: 25 February 2022 09:16 To: Ed .<mailto:ej...@hotmail.com>; David Mertens<mailto:dcmertens.p...@gmail.com> Cc: pdl-devel<mailto:pdl-devel@lists.sourceforge.net> Subject: Re: [Pdl-devel] PDL 2.075 released Hi Ed, I do not mind the behaviour as such but rather the fact that it changed. That breaks previous code in non-obvious ways. There are so many functions which either take a piddle or perl scalar/list as input - or not. The empty piddle case, for example, changed the result from undef to BAD, which in perl context changes from false to true. This may or may not through errors or worse change the result of old code. I don't know what others think but would you mind reverting to the old behaviour? Ingo On 2/24/2022 8:13 PM, Ed . wrote: Hi Ingo, You can get exactly the old behaviour if you were to use “max(which($pos($c_loc)>-1))->at”, since that’s the bit that was removed from the Ufunc code. Best regards, Ed From: Ingo Schmid<mailto:ingo...@gmx.at> Sent: 24 February 2022 19:08 To: Ed .<mailto:ej...@hotmail.com>; David Mertens<mailto:dcmertens.p...@gmail.com> Cc: pdl-devel<mailto:pdl-devel@lists.sourceforge.net> Subject: Re: [Pdl-devel] PDL 2.075 released Ed & David, I stumbled across the problem because of this behaviour: pdl> p max pdl[] BAD which does not expand to 0 as would an undef perl scalar. Here's the function. The isbad() line was necessary to catch the case where which returns an empty list. sub openBox { my $filename=shift; my $filename2=shift; my $time=shift; $c_ti = 0; $c_loc = 0; if ( -f $filename && -f $filename2) { $c_loc = 4; $pos=rcols( $filename, [], {LINES=>"1:"})->transpose; # { HEADER => "$border{0} $border{1} $border{2} $border{3}"}; $posSF=rcols( $filename2, [], {LINES=>"1:"})->transpose; # { HEADER => "$border{0} $border{1} $border{2} $border{3}"}; do { $c_loc--; $c_ti=sclr max (which ($pos($c_loc)>-1)); $c_ti=0 if isbad($c_ti); #say "border $c_loc time $c_ti"; } while ($c_ti <= 0 && $c_loc>0) ; #say "Reading from file $filename; Size of box file ",$pos->info; } else { $pos = zeroes(4,$time)-1; $posSF = zeroes(4,$time)-1; } #say "border (open) $c_loc"; } On 2/24/2022 5:51 PM, Ed . wrote: Hi David, Can you show a complete (small) example of using the return value of `max` in Boolean context that fails with the current code? I am unable to reproduce this, and it ought to work correctly. The change was made in Aug 2021, with version 2.056. Those functions were never documented as returning Perl scalars, and I was extremely surprised when I discovered it then did so. Best regards, Ed From: David Mertens<mailto:dcmertens.p...@gmail.com> Sent: 24 February 2022 15:53 To: Ingo Schmid<mailto:ingo...@gmx.at> Cc: pdl-devel<mailto:pdl-devel@lists.sourceforge.net> Subject: Re: [Pdl-devel] PDL 2.075 released To follow-up on max/min being piddles, this broke some of my tests in PDL::Parallel::threads. Specifically croak-in-boolean-context is now easily triggered where it used to be safe: if ($data->max < 40) { # croaks: piddle in boolean context .... } Was this use case considered and discussed? I missed it, though I haven't always paid terribly close attention. David On Thu, Feb 24, 2022 at 10:18 AM Ingo Schmid <ingo...@gmx.at<mailto:ingo...@gmx.at>> wrote: HI Ed, here's some feedback to the new release. Thank you for your tireless efforts! I experience errors in previously working code, at least in 2.074 and 2.075. $r and $i are piddles, as are $empty_lines and $vlines. These lines (67 #!perl use PDL; use PDL::NiceSlice; use 5.10.0; use strict; sub foo { my $r=shift; #piddle my $i=shift; #piddle my $empty_lines=(whichND ($r(0,,0,0,0,0,0,0;-)==0)); unless ($empty_lines->isempty) { $empty_lines=$empty_lines(0,;-) if $empty_lines->nelem; my $vlines=$$p{y}-1-$empty_lines; $r(,$empty_lines,).=$r(,$vlines,); # syntax error $i(,$empty_lines,).=-$i(,$vlines,); # syntax error } } syntax error at /home/ingo/perl/test_errors.pl<http://test_errors.pl> line 17, near "$r(" syntax error at /home/ingo/perl/test_errors.pl<http://test_errors.pl> line 18, near "$i(" Could that be a PDL::NiceSlice issue? Also, max (and probably min) are piddles now, (tested in 2.074), which breaks some code. They used to returen perl scalars. On 2/19/22 10:22, Ed . wrote: Dear PDL folks, PDL 2.075 has just been released. Notable changes since 2.064: 1. threading now called broadcasting (compat aliases created) 2. fix when readdata functions return error (#356) 3. RedoDimsCode can now use $SIZE(other_index) (#386) 4. can now unify Code and BadCode into Code, and use PDL_IF_BAD() or #ifdef PDL_BAD_CODE 5. add macros PDL_IF_GENTYPE_{REAL,INTEGER,UNSIGNED}(iftrue,iffalse) in Code 6. fixes to PDL::NiceSlice including performance bug by monkey-patching Text::Balanced::extract_multiple 7. Minuit and Slatec 64-bit safe 1. bifuncs like Ops::plus no longer need explicit swap parameter Future plans, in something like intended order: 1. fix more open GitHub issues 2. make PDL::LinearAlgebra work right with “native complex” (several of the above issues were related to enabling this) 3. “loop fusion” techniques to maximise locality of computation, minimising data’s trips through the “straw” between CPU and main RAM 4. finish the independent C interface for making PDL usable from e.g. Python 5. use OpenCL or other means to also utilise GPUs if available The IRC channel (#pdl on irc.perl.org<http://irc.perl.org>) is a great virtual place to come and ask questions, or just watch the GitHub messages flow by. As usual, please give the new PDL a try and report problems. Best regards, Ed _______________________________________________ pdl-devel mailing list pdl-devel@lists.sourceforge.net<mailto:pdl-devel@lists.sourceforge.net> https://lists.sourceforge.net/lists/listinfo/pdl-devel _______________________________________________ pdl-devel mailing list pdl-devel@lists.sourceforge.net<mailto:pdl-devel@lists.sourceforge.net> https://lists.sourceforge.net/lists/listinfo/pdl-devel -- "Debugging is twice as hard as writing the code in the first place. Therefore, if you write the code as cleverly as possible, you are, by definition, not smart enough to debug it." -- Brian Kernighan
_______________________________________________ pdl-devel mailing list pdl-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/pdl-devel